Google ecosystem newby - Can you train an Ollama agent on the Google cloud?

Google AI Studio
, ,
1

Hi All,
As stated, I have only just looked at moving everything on to the cloud from my local lab, so please be patient with my understanding of the ecosystem.

Currently I have a running Ai agent model that produces simple games from user based text inputs, no user coding experience required. The agent takes the users description of the game and produces all coding for physics, graphics, etc., then just spits out the source files and an exe file of the game.

Currently, it is in very basic form, with my current hardware its training period appears it may take days. Seeing we are in start-up, I would prefer not to spend a fortune on Nvidia Ai cards so was looking to utilise the Google cloud by running my training.

Is there anyway to be able to achieve uploading my agent progress based on the Ollama LLM and having it trained on the cloud?

2 Likes
3

Hi @Jebus Welcome to the community
Yes, you can definitely train your AI agent’s underlying LLM on Google Cloud!

While Ollama helps you run models locally, for the actual training or fine-tuning:

  1. You’ll upload your training dataset and your training scripts to Google Cloud Storage

  2. You’ll then use Vertex AI Training . This service lets you run your training script on powerful cloud GPUs.

  3. Once training is complete, Vertex AI saves your fine-tuned model (your “agent progress”) back to GCS.

  4. You can then download this improved model. To use it with Ollama locally, you might need to convert it to a compatible format .

This is the high-level overview of the process which you can try for your usecase
Thank You